JobMatchers — Autonomous Job-Application Agent
A B2C platform that discovers, scores, tailors, submits, and tracks job applications on behalf of career-changers and immigrants. Built for the people the job market under-serves.
The Problem
The job market punishes people who are pivoting. Career-changers and immigrants get filtered out by ATS systems calibrated for “linear” résumés, then drown applying to 200 jobs by hand for a 1% response rate. The canonical user — a friend named Oksana — was sending 30 applications a week and hearing back from one.
A reasonable response is “rewrite your résumé.” A better one is to automate the entire job-search pipeline so the human only sees the matches that actually deserve attention.
The Architecture
End-to-end pipeline: Discover → Score → Tailor → Submit → Track.
- Discover — multi-source job scraping across LinkedIn, Indeed, and ATS direct feeds. Deduplicated, normalized, freshness-tracked.
- Score — fit scoring against the user’s profile, target roles, and stated constraints. Each match is rated and explained — not a black box.
- Tailor — résumé and cover-letter generation per job, conditioned on the specific job description. Edits, not rewrites — the user’s voice survives.
- Submit — autonomous application submission with human-in-the-loop approval on the first few, then trusted autonomous after the user confirms quality.
- Track — single-timeline view of every application, status changes pulled from email parsing, follow-up scheduling, response analytics.
Quarterly prepay subscription model. Single-timeline UX — no Trello-style boards, no kanban dashboards, just one chronological feed of what’s happening.
Key Decisions
Why ICP is career-changers and immigrants, not “everyone looking for a job.” The market rewards focus. Career-changers and immigrants are the segment where the existing job-search products fail hardest — and where automation creates the most leverage per dollar. A generalist tool serves nobody well; a specialist tool builds word-of-mouth in a tight community.
Why quarterly prepay, not monthly. Job searches don’t fit a monthly cycle. They take 8-16 weeks. Quarterly prepay aligns price with outcome and filters for serious users who actually need the tool to work.
Why single-timeline UX. Job-seekers already feel out of control. The last thing they need is another dashboard demanding triage. One feed, chronological, actionable items only.
Why human-in-the-loop on early submissions. Trust gets earned. The agent submits the first applications with the user approving each one; once the user has approved five in a row, the system can run autonomously. The user sets the pace, not the product manager.
What I Learned
-
The AI is not the hard part. The hard part is the data pipeline — keeping job listings fresh, deduplicating across sources, handling broken ATS forms, parsing inconsistent response emails. The agent is 20% of the work.
-
Specificity wins B2C. Generic “AI résumé builders” have ten thousand competitors. “Autonomous job-application agent for immigrants” has almost none. The narrower the ICP, the easier the marketing.
-
Career-change customers are different. They’re not optimizing for the best job — they’re optimizing for any job that lets them survive the transition. The product has to take that emotional weight seriously. Treating it like a B2B tool would be a category error.
Status
In active build as of May 2026 in ~/projects/archive/jobhunter-precursor/ (despite the path — the brand has evolved from JobHunter to JobMatchers, repo already renamed). Domain: www.jobmatchers.app.